Health Care Spending and Hidden Poverty in India by Michael Keane University of New South Wales Ramna Thakur Indian Institute of Technology, Mandi August 3, 2018 Abstract India has a high level of out-of-pocket (OOP) health care spending, and lacks well developed health insurance markets. As a result, official measures of poverty and inequality that treat medical spending symmetrically with consumption goods can be misleading. We argue that OOP medical costs should be treated as necessary expenses for the treatment of illness, not as part of consumption. Adopting this perspective, we construct poverty and inequality measures for India that account for impoverishment induced by OOP medical costs. For 2011/12 we estimate that 4.1% of the population, or 50 million people, are in a state of “hidden poverty” due to medical expenses (based on official poverty lines). Furthermore, while poverty in India fell substantially from 1999/00 to 2011/12, the fraction of the remaining poverty that is due to medical costs has risen substantially. Economic growth appears less “pro-poor” if one accounts for OOP medical costs, especially since 2004/05, and especially in rural areas. Finally, we look beyond poverty rates to show how OOP health costs affect the entire shape of the consumption distribution. JEL Codes: I14, I32, O15 Keywords: Poverty, Consumption, Healthcare, Medical Costs, Inequality, Growth. Acknowledgements: Support for this project was received from the EU’s Erasmus Mundus Experts4asia program, and from Australian Research Council grant FL110100247. I. Introduction In the 70 years since independence, India experienced substantial improvements in life expectancy, and both infant and maternal mortality. Yet health outcomes still lag behind many other developing countries (see Alkire and Seth, 2015). The Indian government’s expenditure on health care is only about 1% of GDP, and less than 16% of the population is covered by health insurance (WHO, 2006, MOSPI, 2015). Thus, India’s health system ranks as one of the most heavily dependent on out-of-pocket expenditures in the world. According to GOI (2005) and WHO (2006), roughly 70% of health expenditure in India is out-of-pocket (OOP). In this paper, we examine the impact of OOP health expenditures on the incidence and depth of poverty. We argue that official measures of poverty derived from consumption surveys tend to mask poverty due to OOP health care costs. An adverse health shock S has the effect of shifting the expenditure function e(U,S) upward, raising the expenditure level e required to attain any given level of utility U. This means the relevant poverty line for a person who suffers a health shock S is shifted upward by (at least) the cost of treating the illness.1 From this perspective, a person with consumption C who suffers an episode of illness that requires R rupees to treat is no better off than a comparable but healthy person with consumption of only C-R. Letting P denote the poverty line, we define “hidden poverty” (C>P but C-R<P) and “reinforced poverty” (C<P, C-R<C<P). We then use consumer expenditure survey (CES) data from the National Sample Survey Organization (NSSO) to construct new poverty measures for India that account for impoverishment either induced or reinforced by medical costs. Specifically, we use the NSS-CES data for the period 1999-2000 (55th round), 2004-05 (61st round) and 2011-12 (68th round). The use of three surveys allows us to examine trends in OOP spending and its effect on poverty calculations and on inequality measures. Prior studies on the effect of health care costs on poverty in India obtain a wide range of results. Peters et al. (2002) use the NSS Health Survey for 1995-96 and estimate the poverty rate would increase by 2.2 percentage points if one accounts for OOP costs. Garg and Karan (2009) argue the Health Survey data understate OOP costs, so they use NSS-CES data for 1999-2000. They estimate a 3.2 point increase in the poverty rate due to OOP costs. Van Doorslaer et al. (2006) also use the NSS-CES data for 1999-2000, and estimate that accounting for OOP costs would increase the poverty rate by 3.7 points (from 31.1% to 34.8%). This corresponds to 37 1 Of course, illness may have other costs besides treatment costs (e.g., pain and suffering). So in general e(U,S) may shift up by even more than the cost of treatment. We abstract from this issue, which means we are likely to understate the true cost of health shocks. 1 million people in hidden poverty. In contrast, Ravi et al. (2016) use the 60th and 71st NSS rounds for 2004 and 2014. They estimate that accounting for OOP costs increases the poverty rate by 7.5 points – at least twice as large as affects found in the earlier studies. This suggests the extent of hidden poverty in India due to OOP medical costs is controversial. Notably, the 60th and 71st NSS were “health rounds” designed to measure utilization of health services, not health spending or total consumption. Indeed, NSSO (2006) pp. 10-11 states these data underestimate total consumption. Thus, they overestimate the ratio of OOP costs to total consumption and exaggerate hidden poverty.2 This is why we limit our analysis to the NSS- CES data for 55th, 61st and 68th rounds, which are the three most recent “consumption rounds” designed to provide accurate data on both health spending and total consumption. Aside from using the best recent data, we extend earlier analysis in other important ways. First, we update results to 2012, and look at trends in poverty and inequality from 1999/2000 to 2011/12. India experienced historically rapid economic growth in the 2000s, so it is important to document how this affected poverty rates and income distribution, and to gauge the impact of OOP health costs in this period. This aspect of our paper is related to work by Fosu (2017) and Ravallion (2001) who study how economic growth translates into poverty reductions. Fosu (2017) shows India was relatively inefficient at converting economic growth into reduced poverty in the 1994-04 period. For 80 developing countries, he finds India’s elasticity of the poverty rate with respect to income was only about half the median value. We ask whether growth in India became more “pro-poor” in the 2005-2012 period, and, if not, whether India’s poor performance may be due (in part) to rising OOP health care costs. Second, we analyze the effect of OOP costs on the entire consumption distribution, not just individual statistics like the poverty rate or poverty gap. We discuss how poverty rates can be misleading, and extend earlier work by looking at the differential impact of OOP health costs at different quantiles of the consumption distribution. Finally, we also examine how accounting for health care costs affects traditional inequality measures like the Gini, and show how these scalar measures can be quite misleading regarding impacts of OOP on inequality. Our calculations show the poverty rate is substantially understated by failure to account for OOP health care costs. For instance, in the 2011-12 survey we estimate a poverty rate of 2 By comparing Tables 2 and 12 in Ravi et al. (2016), one can verify that their figures imply the health spending share of consumption in India was 10.7% in 2004 and 11.9% in 2014. These figures are implausibly high for any developing country – in fact, they are as high as US levels. Our figures (see Figure 1) are 5.6% for 2004/5 and 6.8% for 2011/12, which are much more reasonable (and roughly in line with India national accounts figures). 2 22.3% under the standard total consumption measure, rising to 26.4% if we deduct OOP costs. This increase of 4.1% represents roughly 50 million people (as the population of India in 2012 was 1.22 billion). Thus, our estimate of hidden poverty due to medical costs is higher than earlier estimates (except for the Ravi et al (2016) results which we consider implausibly high). Our trend analysis shows that the conventionally measured poverty rate fell from 42.3% to 37.2% between the 1999/2000 and 2004/5 surveys, and then to 22.3% in the 2011/12 survey. Once we factor in OOP medical costs, the figures are 46.9%, 30.5% and 26.4% respectively. Thus, by any standard, poverty fell substantially in India over this period. The decline was 47.3% if we look at the conventional poverty rate measure. However, if we factor in hidden poverty due to medical costs, the decline was a bit more modest at 43.7%. It is important to view these large poverty reductions in the context of the stunning growth in per capita consumption that India achieved over this period (i.e., 3.44% annually from 1999/2000 to 2004/05, and 6.65% annually from 2004/05 to 2011/12. The poverty reduction in 1999-2004 was actually modest in light of the rapid growth rate (consistent with Fosu, 2017). Our results show India’s growth became only slightly more “pro-poor” in 2005-12. Hidden poverty due to medical costs rose from 10.9% of measured poverty in 1999/2000 to 18.2% in 2011/12. Thus, as poverty rates have fallen, medical costs have become a larger factor contributing to current poverty. The poor state of the health care system (see Section II) may be a key reason that economic growth in India has not been more “pro-poor” (see Ravallion and Datt, 2002).